Friday, March 19, 2010

Can Polling Memos Change the Narrative About 2010 Races?

By Stuart Rothenberg

Here’s a bulletin for you: Anytime a campaign releases a polling memo, it is making an argument, not merely offering survey data for your information. Polling memos aren’t written to make you smarter.

This shouldn’t need to be said, of course, but when I see reporters swallowing spin as if it were information — as one Louisiana Gannett reporter did recently in writing about the state’s Senate race — I get very uncomfortable.

A number of Democratic polling memos from reputable polling firms have been circulating over the past couple of weeks — two from Anzalone Liszt Research about the Louisiana Senate race and about Rep. Bobby Bright’s (D-Ala.) re-election prospects, and one from Harstad Strategic Research about the Colorado Senate race — and readers should understand what’s going on with them.

I am not, I must emphasize, challenging the data. These are credible polling firms, and almost every pollster I know has released these kinds of memos in the past. I am only using Anzalone Liszt and Harstad as examples.

The purpose of the Louisiana poll memo is to alter the developing narrative that the state’s Senate race is essentially over and that Sen. David Vitter (R) won’t be seriously threatened by Rep. Charlie Melancon (D). The Alabama memo seeks to create a sense of inevitability about Bright’s re-election prospects and undercut GOP challenger Martha Roby’s credibility (and fundraising and buzz).

The Colorado memo seeks to rebut polling that shows appointed Sen. Michael Bennet (D) trailing in his race and to help him build momentum for his primary and the fall election.

I recently asked a pollster, not for attribution of course, about these kinds of memos and received a gloriously forthright answer: “Anyone who does a survey for strategic reasons isn’t going to release strategic information.”

Pollsters say that their surveys present only a “snapshot” of a race at a particular moment. That’s true. But often the snapshot presented in a memo is misleading, and the pollster knows it. Memos include numbers intended to build an argument that seems empirically based but isn’t. They don’t present the whole picture, because the whole picture isn’t in their client’s interest.

Even highly regarded, methodologically legit pollsters tell me to call them up privately if I want to get their real assessment about a race — don’t go by the memo they release. I get this from both Republican and Democratic pollsters, and I have received the same advice for years.

Actually, most pollsters hate to write these kinds of memos, but their clients want them to create a more favorable narrative, so they write them, usually using their words very carefully.

The Bright memo begins “Congressman Bobby Bright is well positioned to win re-election in Alabama’s 2nd Congressional district. Bright’s personal popularity and positive job rating are extremely high.” The memo looks at the Congressman’s excellent favorable/unfavorable ratings, respondents’ answers to questions about Bright’s qualities and his leads of 24 to 32 points over possible November opponents.

The memo ends with the following paragraph: “Bright leads Roby among virtually all gender, race, and geographic subgroups. Bright earns double-digit margins with white voters, African American voters, Independent voters, and in both the Montgomery and Dothan media markets.”

Bright, a conservative Democrat, won election to Congress because of the Democratic wave in 2008. He beat a politically damaged GOP nominee who had emerged from a bruising primary by six-tenths of a point (50.2 percent to 49.6 percent), winning by 1,790 votes out of more than 286,000 cast.

However highly regarded Bright is, this is a Republican district. President Barack Obama drew only 37 percent of the vote in it in 2008, losing it to Sen. John McCain (R-Ariz.) by more than 25 points.

Bright does start off well-positioned for his re-election bid, but he is almost guaranteed to have a razor-close race against Roby, a Montgomery city councilwoman and the likely GOP nominee. Maybe he’ll win; maybe he won’t. Whatever the outcome, he won’t win by 20 points or carry white voters by double digits.

If Bright holds on to win re-election, he’ll win by a point or two. Given that, the Anzalone Liszt poll tells us little about November. Voters apparently don’t know anything about the Republican candidates, and it’s so far from Election Day that most voters haven’t given much thought to what their vote might mean or who they really will vote for in the fall.

The Bright memo reminds me of a June 15, 2009, Anzalone Liszt polling memo that listed all of Democrat Creigh Deeds’ advantages in the 2009 Virginia gubernatorial race.

That memo asserted: “Deeds has a high favorable rating, and a lower unfavorable rating, even though McDonnell spent more on television [during the primary]. Deeds also holds critical issue advantages that will make it difficult for McDonnell to make up ground.”

The memo said that Deeds held a 4-point lead in the race and was able to win votes across the state “including traditional Republican strongholds.” “At the same time, McDonnell will have a difficult time making inroads in increasingly Democratic Northern Virginia,” it asserted.

Of course, McDonnell won 59 percent to 41 percent, turning a 4-point deficit in the poll into an 18-point victory. Deeds didn’t do well in “Republican strongholds,” and McDonnell carried Fairfax County, the state’s largest county and the epitome of what is meant by “Northern Virginia.”

Polling memos sometimes contain useful nuggets of data, but they often leave out other important data and stress the narrative the campaign wants to create. Don’t take them at face value.

This column first appeared in Roll Call and on on March 18, 2010. 2010 © Roll Call Inc. All rights reserved. Reprinted with permission.